CODEWITHSUNDEEP
raggregate
S.K.
S.K.

Reputation: 365

Aggregate Columns by weighted average or simple average if weight is missing

I am trying to aggregate certain columns of a data frame. In my data frame each column correponds to an industry and each row to a particular country. Idealy I would like to aggregate certain columns by weighted average. However for a small fraction the weights are missing. In this case I would like that R would aggregate for this country the industries by a simple average. A snipet from the data frame (these are exemplary weights for other columns) 

|   Mining   | Food     |  weight85| weight90.93|
|:----------:|--  -----:|---------:|------------|
|   0.9608709| 0.8839236| 0.2738525|   0.1943577|
|   0.6445055| 0.8483874| 0.2958678|   0.1043844|
|   0.6977353| 0.9449249|        NA|          NA|
|   0.7970192| 0.5941056| 0.2324452|   0.1904089|
|   0.7261323| 0.6333187|        NA|          NA|
|   0.9959837| 1.0101725| 0.3872314|   0.1628354|

I compute the weighted average when ingoring the missing values problem as follows:

  GGPC$mining.weighted <- GGPC$weight85*GGPC$Mining
  GGPC$food.weighted  <- (1-GGPC$weight85)*GGPC$food
  GGPC$food.mining<- rowSums(GGPC[,54:55], na.rm=T)

Upvotes: 1

Views: 239

Answers (1)

S.K.
S.K.

Reputation: 365

Building on the answer which mts provided. I came up with following solution for a solution, which computes for one row either the simple average or the weighted average. 

 if(sum(is.na(DF[1,37])>0)) {1/2*DF[1,5]+1/2*DF[1,6]}
 else  { DF[1,37]*GGPC[1,5]+(1-DF[1,37])*DF[1,6]}

And further looping through the rows of a data frame

 DF$data.column.agg <- 0
 for (i in 1:length(DF)) {
  DF[i,*data.column.agg*] <- if(sum(is.na(DF[i,*weight column*])>0))  {*simple average* }     
 else {DF[i,*weight column*]*GGPC[i,*data column1*]+(1-DF[i,*weight column*])*GGPC[i,*data column2*]} 
 }

Upvotes: 0

Related Questions